ORIGINAL RESEARCH
published: 11 June 2019
Edited by:
Ashish Raj,
University of California,
San Francisco, United States
Reviewed by:
Thomas Koenig,
University of Bern, Switzerland
Scott Edward Counts,
Michigan State University,
United States
*Correspondence:
Christian Sandøe Musaeus
christian.sandoee.musaeus@
regionh.dk
Specialty section:
This article was submitted to
Neurodegeneration,
a section of the journal
Frontiers in Neuroscience
Received: 25 February 2019
Accepted: 15 May 2019
Published: 11 June 2019
Citation:
Musaeus CS, Nielsen MS and
Høgh P (2019) Microstates as
Disease and Progression Markers
in Patients With Mild Cognitive
Impairment. Front. Neurosci. 13:563.
doi: 10.3389/fnins.2019.00563
Frontiers in Neuroscience | www.frontiersin.org
doi: 10.3389/fnins.2019.00563
Microstates as Disease and
Progression Markers in Patients With
Mild Cognitive Impairment
Christian Sandøe Musaeus 1* , Malene Schjønning Nielsen 2 and Peter Høgh 2,3
1
Department of Neurology, Danish Dementia Research Centre, Rigshospitalet, University of Copenhagen, Copenhagen,
Denmark, 2 Regional Dementia Research Centre, Department of Neurology, Zealand University Hospital, Roskilde, Denmark,
3
Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark
Network dysfunction is well established in patients with Alzheimer’s disease (AD) and
has been shown to be present early in the disease. This is especially interesting in
patients with mild cognitive impairment (MCI) since they are more likely to develop
AD. In EEG, one type of network analysis is microstates where the EEG is divided
into quasi-stable states and these microstates have been linked to networks found
with resting state functional MRI. In the current exploratory study, we therefore wanted
to explore the changes in microstates in MCI, and AD compared to healthy controls
(HC) and whether microstates were able to separate patients with MCI who progressed
(pMCI) and those who remained stable (sMCI). EEGs were recorded at baseline
for 17 patients with AD, 27 patients with MCI, and 38 older HC and the patients
were followed for 3 years. To investigate whole-brain dynamics we extracted different
microstate parameters. We found that patients with MCI, and AD had significantly
higher occurrence (p-value = 0.028), and coverage (p-value = 0.010) for microstate
A compared to HC. However, we did not find any significant systematic deviation of the
transition probabilities from randomness for any of the groups. No significant differences
were found between pMCI and sMCI but the largest difference in duration was found for
microstate D. Microstate A has been linked to the temporal lobes in studies combining
EEG and fMRI and the temporal lobes are the most affected by AD pathology in the early
stages of the disease. This supports our idea that microstate A may be the first affected
microstate in early AD. Even though not significant between pMCI and sMCI, Microstate
D has previously been shown to be associated with both frontal and parietal areas
as measured with fMRI and may correspond to underlying pathological changes in the
progression of MCI to AD. However, larger studies are needed to confirm these findings.
Keywords: EEG, mild cognitive impairment, Alzheimer, Alzheimer’s disease, progression, stable, MCI, microstate
1 June 2019 | Volume 13 | Article 563
Musaeus et al. EEG Microstates in MCI

INTRODUCTION MATERIALS AND METHODS

Alzheimer’s disease (AD) is a progressive neurodegenerative Recruitment, Inclusion Criteria, and
disease and patients with AD have shown changes in functional
Subjects
brain networks (Dickerson and Sperling, 2009). Studies have
The whole dataset or parts of the dataset have also been used
even suggested that alterations in networks are present very
for other studies (Engedal et al., 2015; Musaeus et al., 2018a,
early in the disease process (Selkoe, 2002; Cummings, 2004).
2019; Nielsen et al., 2018) including separate presentation of
Patients mild cognitive impairment (MCI), which is thought
results from spectral power analysis (Musaeus et al., 2018b)
of as mild objective cognitive deficits, are associated with
and for coherence and imaginary part of coherency analysis
later development of AD (Petersen et al., 1999; Petersen,
(Musaeus et al., 2019).
2004). While some patients with MCI progress (pMCI)
This prospective cohort study was conducted at two
others remain stable (sMCI) in their disease, which is in
Danish memory clinics at Zealand University Hospital and
large due to different etiological causes as for example
Rigshospitalet, respectively. Patients consecutively referred for
depression or vascular changes. However, for patients with
cognitive evaluation and diagnosed with either MCI or mild AD
MCI due to AD, there is also evidence of fast and slow
and at least a baseline Mini-Mental State Examination (MMSE)
progression (Chui, 1987), which may be due to affection of
score of ≥22 were eligible for inclusion. The patient selection
different brain networks.
was defined using preexisting exclusion criteria: (1) no close
Multiple methods to investigate brain networks have
relatives who wished to participate, (2) if they were participating
been proposed with the most common being fMRI. But
in other intervention studies or (3) if they were suffering from
since network function are thought of as fast processes
other neurological, psychiatric, or other severe disease, (4) if
that changes over time, fMRI may not be able to capture
they received sedative medication due to a potential sedative
these. Electroencephalography (EEG) has a high temporal
effect, and (5) if they had any past or current addictions to
resolution and methods like microstate analysis has been
alcohol or medications.
able to show topographical maps that have been associated
The HC were all volunteers recruited trough public
with resting state networks (Van de Ville et al., 2010;
advertisements at the memory clinics, at local associations
Yuan et al., 2012). Microstates is a technique where the
for elders and through an online recruitment site for trial
multichannel resting-state EEG signal can be divided
subjects. Inclusion criteria were: (1) age between 50 and
into a number of distinct states (Lehmann et al., 1987).
90 years, (2) MMSE score ≥26, (3) ACE ≥85, (4) normal
Although these states occur in a time range of milliseconds
neurological and clinical examination, (5) normal or age-related
(ms), it has been shown that momentary stable spatial
brain atrophy measured on a computed tomography (CT)
patterns that last approximately 80–120 ms before rapidly
scan, (6) normal routine blood tests. Exclusion criteria
transitioning to a different microstate (Khanna et al., 2015).
were: (1) an inability to participate (including impaired
The majority of the studies have clustered the resting
vision or hearing), (2) presence of cognitive symptoms
EEG into four microstate classes, which has been found
including memory complaints, (3) signs of major neurological,
to be the optimal number according to cross-validation
psychiatric or other severe disease, which potentially could
criterion (Pascual-Marqui et al., 1995; Koenig et al.,
elicit cognitive impairments including signs of major
2002) and a study found a high test-retest reliability
depression or a geriatric depression scale score >7, (4) be
(Khanna et al., 2014).
pregnant, (5) have undergone general anesthesia, (6) received
Only few studies have investigated alterations in microstates
electroconvulsive therapy in the past 3 months, (7) receive
in patients with AD (Ihl et al., 1993; Dierks et al., 1997;
sedatives, or (8) have any past or current addictions to
Strik et al., 1997; Stevens and Kircher, 1998; Nishida et al.,
alcohol or medications.
2013). Most studies found a shorter duration of the microstates
In total, we included 17 patients with AD, 27 patients with
compared to healthy elderly controls with one study finding
MCI, and 38 HC. The study was reported to and approved by
a longer duration of microstates. However, a more recent
the Danish Data Protection Agency and by the Regional Ethical
study has found no significant changes in either duration or
Committee according to Danish legislation.
occurrence in patients with AD compared to healthy controls
(Nishida et al., 2013). Moreover, none of the studies have
investigated the early changes in microstates by looking at Diagnostic Assessment
patients with MCI or whether microstates are able to differentiate The patients underwent a standardized diagnostic assessment
between pMCI and sMCI. including a full physical and neurological examination, routine
In the current exploratory study, we wanted to investigate blood analysis, brain CT or MRI scan as well as cognitive
the changes in microstates in patients with MCI compared to screening, i.e., MMSE, Addenbrooke’s Cognitive Examination
both AD, and healthy controls (HC). Furthermore, we wanted (ACE), Digit Symbol Substitution Test (DSST), and Clinical
to investigate whether microstates can be used to separate Dementia Rating (CDR). Furthermore, as part of the diagnostic
pMCI from sMCI. Lastly, we wanted to investigate whether any assessment patients and relatives underwent NeuroPsychiatric
microstates parameters correlated with either cognitive scores Inventory (NPI), Major Depression Inventory (MDI), Activities
or AD biomarkers. of Daily Living Inventory (ADCS-ADL). The CT and MRI

Frontiers in Neuroscience | www.frontiersin.org 2 June 2019 | Volume 13 | Article 563

Musaeus et al.
scans were examined by a neuro-radiologist. The majority
also had a lumbar puncture (except two patients with
MCI and six HC) performed to measure AD biomarkers
(Amyloid-β42 , total tau, and phosphorylated tau), and for
routine parameter analysis. If diagnostically relevant, the patients
also had a neuropsychological evaluation undertaken by a
clinical neuropsychologist, but these were individualized for
each patient with varying overlap and therefore not included
in the current study. Diagnoses were settled by consensus of
a multidisciplinary team based on all examination results. The
included MCI patients fulfilled the Winblad consensus criteria
(Winblad et al., 2004) and AD patients fulfilled the NIA-AA
criteria (McKhann et al., 2011).
At inclusion, all HC underwent the standardized diagnostic
assessment, which included cognitive tests (ACE, MMSE, DSST),
MDI and analysis of CSF was performed on almost all HC. At the
baseline visit all HC were referred for a standardized EEG. The
EEG recordings were not used in the diagnostic assessment.
Study Design
The patients were recruited within 6 months after the diagnosis
and all tests were repeated at inclusion. Follow-up visits were
carried out on a yearly basis, with serial cognitive tests, i.e.,
MMSE and ACE and the NPI, MDI, ADCS-ADL, and CDR
scales. Clinical progression of MCI to AD was determined
based on whether the patient clinically fulfilled the NIA-AA
criteria (McKhann et al., 2011). If the patient progressed to
another diagnosis, they were excluded from the comparison
between pMCI and sMCI.
The primary investigator performing the tests was blinded for
the results of the EEG, imaging and CSF analysis during the study
period. This was done for the investigator to be blinded for the
potential presence of underlying AD pathology.
Electroencephalography Recording
The EEG recordings were performed at the two participating
centers and the EEG recordings were performed using
NicoletOne EEG Systems (Natus
) with a sampling rate of
R
either 500 or 1000 Hz. Nineteen electrodes were positioned
according to the International 10–20 system. Most EEGs were
recorded with alternating eyes closed (EC) and eyes open periods
for 3 min each but some of the recording only had EC segments.
The participants were alerted if they became visibly drowsy, since
drowsiness influences recording. The neurophysiology assistant
recording the EEG made marks in the EEG when the participant
closed and open their eyes. After the recording, the files were
exported as raw EEGs without any filtering.
Collection and Analysis of Cerebrospinal
Fluid
The lumbar puncture was performed between the L3/L4 or L4/L5
intervertebral space and the CSF was collected in polypropylene
tubes. Analysis of the CSF included routine parameters and
the core AD biomarkers, i.e., Aβ42 , T-tau, and P-tau. The AD
biomarkers were quantified with sandwich ELISAs [INNOTEST
amyloid-β42 , hTau, and Phospho-Tau (181P), respectively;
Frontiers in Neuroscience | www.frontiersin.org
EEG Microstates in MCI
Fujirebio Europe, Ghent, Belgium]. AD biomarkers analyses
from both clinics were all carried out at one central laboratory.
Preprocessing of EEG
The EEG data were imported to MATLAB (Mathworks, v2016a)
using the EEGLAB toolbox (Delorme and Makeig, 2004). Only
segments with EC were selected either using markers placed
doing recording or from the first 10 min of recording if markers
were not present. The electrodes were computationally located on
the scalp using the dipfit toolbox (Oostenveld et al., 2011) with
the standard 10–20 electrode model. The excessive channels were
removed, and the data were bandpass filtered from 1 to 70 Hz
using the pop_firws function in MATLAB with a filter order of
2 and the Kaiser window parameter beta was estimated using a
maximum passband ripple of 0.001. Furthermore, the data were
bandstop filtered from 45 to 55 Hz using the same settings as
described previously. Afterward, the data were down sampled
to 200 Hz. Then, the data were divided into 1 s epochs and the
EEGs were visually inspected and epochs with excessive noise or
artifacts were removed. Channels with excessive noise, drift, or
bad connection were interpolated using spherical interpolation.
The EEG had to have ≤ three electrodes with excessive artifact,
otherwise the EEG was excluded from the analysis. Afterward, the
EEGs were re-referenced to average and independent component
analysis (ICA) was performed using the extended infomax
algorithm (Lee et al., 1999) for each file and components that
contained eye blinks, eye movement, or specific line noise
artifacts were removed manually. Lastly, the EEGs were inspected
visually again and epoch with excessive noise or artifacts were
removed. The investigator who performed the preprocessing was
blinded to the diagnosis. Due to excessive artifacts, we excluded
the following number of EEGs: two from patients with AD, two
from patients with MCI, and one from HC. When comparing
pMCI, and sMCI, one EEG from MCI was excluded due to
clinical progression to vascular dementia.
Microstate Analysis
Before performing the microstate analysis, we first lowpass
filtered the data at 20 Hz with the same settings as mentioned
above. Afterward, we concatenated the epochs for each subject,
i.e., ending up having one continuous EEG file. We performed
the microstates analysis using the Microstate EEGlab Toolbox
(Poulsen et al., 2018). Here, we first extracted the global field
power (GFP) peaks and the settings were a minimum peak
distance of 10 ms, the number of GFP peaks per subject that enter
the segmentation was set at 1000, and GFP peaks that exceeded
2 times the standard deviation of the GFPs of all maps were
excluded. All the GFP peaks from all subjects were aggregated
into one file before segmentation with the goal to maximizing
the similarity between the microstates they would be assigned
to. For segmentation, we used the modified K-means algorithm
since it ignores the polarity of the EEG topography (Lehmann,
1971; Wackermann et al., 1993; Pascual-Marqui et al., 1995).
Here, we predefined the number of microstates as four, which
previously has been reported as the most common (Khanna et al.,
2014) and reproduceable (Khanna et al., 2015). The number of
repetitions were set at 50 and maximum number of iterations
3 June 2019 | Volume 13 | Article 563
Musaeus et al. EEG Microstates in MCI

FIGURE 1 | The global maps that were calculated based the aggregated
dataset from all participants and were back-fitted to each of the EEG
recordings. The labels (A–D) are according to the previous literature in the
field.
were set at 1000. The global maps (see Figure 1) were then
back-fitted to each of the EEG files by labeling each of EEG
segments with the class of microstates it is most familiar. Since
resting state EEG is noisy, it happens that consecutive time
frames are labeled different by change. To avoid this, we rejected
microstate segments shorter than 30 ms. The labels of time frames
in small segments were changed to the next most likely microstate
class, as measured by global map dissimilarity (Poulsen et al.,
2018). After back-fitting the global maps, we calculated global
explained variance (GEV), duration, occurrence, coverage, and
the syntax for EEG files.
As a post hoc examination of the transition probabilities,
we performed the same analysis as previously described in
detail (Lehmann et al., 2005; Nishida et al., 2013). In short, we
calculated the observed transitions based on all transitions and
then the expected transitions based on the occurrence of the
microstates for each subject separately. Afterward, these values
(4 × 4 − 4 = 12) were averaged across subjects for each group,
and the difference was assessed using the chi-square distance. To
statistically test the difference, we performed a permutation test
with 5000 repetitions where the labels “expected” and “observed”
were randomly assigned to the subjects’ sets of the 12 transition
probabilities, and the chi-square distance was computed. The
underlying hypothesis of this test was that if transitions from
one state into the next occurred randomly, observed transition
values would be proportional to the relative occurrence of the
microstate classes.
Statistics
MATLAB (vR2016a) was used for all statistical analyses. When
comparing demographics, number of epochs, average GEV
and cognitive scores for AD, MCI, and HC, we performed
one-way ANOVAs. Independent t-tests were used to compare
baseline cognitive scores between pMCI and sMCI. Since the
microstate results (duration, occurrence, coverage, and syntax)
were non-normally distributed, we log-transformed the data.
Afterward, we performed an ANCOVA with age, gender,
education, and current medication (see Table 1) as covariates.
If we found a significant difference (p-value < 0.05), we
performed independent t-tests (without covariates) between AD
vs. HC, MCI vs. HC, and AD vs. MCI. For the microstate
results from baseline EEG between pMCI vs. sMCI we used an
ANCOVA with the same covariate as mentioned above. The
division of the baseline EEGs into either pMCI and sMCI was
determined on progression after 2nd year follow-up. We used
the data before long-transformation for display in the tables.
For the post hoc analyses between microstate features, we also
calculated effect size measured with Cohen’s d, which is defined
as the difference between two means divided by the standard
deviation of the data.
Furthermore, we performed Spearman’s correlation using
the values from significant differences between AD, MCI, and
HC (coverage, and occurrence for microstate A). We chose to
correlate these values with amyloid, total tau, phosphorylated
tau, MMSE, and ACE.
RESULTS
Demographics, Cognitive Tests, and EEG
Length
Characterization of the patients including cognitive test scores
and EEG length is shown in Table 1. For the performance on
cognitive tests for each visit see Table 2. For the comparison
between demographics, baseline cognitive scores, and CSF
biomarkers for pMCI and sMCI see Table 3. Flow diagram of the
included patients is illustrated in Figure 2.

TABLE 1 | Table showing the characteristics of the participants included in the analysis.

HC (n = 37) AD (n = 15) MCI (n = 25) p-value

Mean age (SD), years 65.7 (6.9) 70.1 (7.8) 71.4 (6.0) 0.006
Female gender, n 17 8 6 0.119
Education, years (SD) 12.7 (3.6) 12.1 (4.0) 10.6 (3.4) 0.105
MMSE, mean (SD) 29.1 (1.0) 26.3 (3.2) 27.6 (1.5) 0.001∗
Antidepressants 1 1 4 0.161
Cholinesterase Inhibitors 0 8 1 0.001∗
Pain killers 2 0 2 0.553
CSF amyloid, mean (SD) 997.5 (320.2) 550.7 (141.2) 782.3 (319.8) 0.001∗
CSF total tau, mean (SD) 303.3 (144.7) 618.4 (186.0) 419.6 (173.9) 0.001∗
CSF phosphorylated tau, mean (SD) 68.5 (103.4) 93.0 (33.3) 59.4 (21.5) 0.384
EEG length, mean seconds (SD) 177.5 (62.1) 147.1 (19.6) 153.6 (44.6) 0.078

HC, healthy controls; AD, Alzheimer’s disease; MCI, mild cognitive impairment; SD, standard deviation; MMSE, Mini-Mental State Examination; CSF, cerebrospinal fluid.
∗ Indicates significant p-value (< 0.05).

Frontiers in Neuroscience | www.frontiersin.org 4 June 2019 | Volume 13 | Article 563

Musaeus et al. EEG Microstates in MCI

TABLE 2 | The cognitive scores, number of participants that dropped out, and number of patients with MCI that progressed doing follow-up for year 2.

Baseline 2nd year follow-up t-value p-value

HC Dropout/total (n) 0 1/37

Progression/no-progression NR NR
MMSE, mean (SD) 29.08 (0.98) 29.36 (0.83) −1.312 0.194
ACE, mean (SD) 94.70 (3.28) 95.58 (3.32) −1.139 0.259
MDI, mean (SD) 3.62 (2.87) 4.06 (3.30) −0.600 0.551
MCI Dropout/total (n) 0 6/25
Progression/no-progression NR 12/13
MMSE, mean (SD) 27.60 (1.50) 26.00 (3.33) 2.138 0.038∗
ACE, mean (SD) 84.13 (8.17) 79.67 (11.59) 1.464 0.151
MDI, mean (SD) 7.13 (5.91) 10.22 (7.75) −1.450 0.155
NPI, mean (SD) 3.38 (3.49) 5.24 (2.49) −1.844 0.073
ADL, mean (SD) 70.71 (4.84) 66.59 (9.87) 1.544 0.133
AD Dropout/total (n) 0 7/15
Progression/no-progression NR NR
MMSE, mean (SD) 26.27 (3.17) 23.50 (5.53) 1.537 0.139
ACE, mean (SD) 77.60 (12.87) 67.14 (18.85) 1.532 0.141
MDI, mean (SD) 5.67 (4.70) 4.17 (4.62) 0.664 0.515
NPI, mean (SD) 1.5 (1.24) 5.00 (2.45) −4.235 < 0.000∗
ADL, mean (SD) 70.86 (8.16) 67.38 (8.67) 0.942 0.358
Missing values (%) 6.78 26.21

In addition, the percentage of missing values for the cognitive scores can be seen. All cognitive scores have been compared over time using a paired t-test. ∗ Indicates
significant p-value (< 0.05).

TABLE 3 | Demographics, baseline cognitive scores, and CSF results for stable mild cognitive impairment (sMCI) and progressed mild cognitive impairment (pMCI).

Baseline – sMCI Baseline – pMCI p-value

(n = 13) (n = 11)

Mean age (SD), years 72.38 (6.06) 70.27 (6.63) 0.424

Female gender, n 4 2 0.500
Education, years (SD) 10.69 (3.84) 10.55 (3.36) 0.922
CSF amyloid, mean (SD) 820.08 (348.64) 695.75 (309.90) 0.419
CSF total tau, mean (SD) 398.25 (162.10) 461.56 (206.29) 0.440
CSF phosphorylated tau, mean (SD) 60.54 (24.54) 59.89 (19.28) 0.948
MMSE, mean (SD) 27.92 (1.38) 27.09 (1.58) 0.182
ACE, mean (SD) 87.54 (6.08) 79.00 (8.36) 0.010∗
MDI, mean (SD) 8.67 (6.89) 6.00 (4.14) 0.297
NPI, mean (SD) 3.09 (3.96) 3.00 (2.24) 0.952
CDR, mean (SD) 0.50 (0) 0.56 (0.17) 0.281
ADL, mean (SD) 70.60 (6.06) 70.86 (2.73) 0.918

T-tests were performed to compare the two groups for each score separately. ∗ Indicates significant (p-value < 0.05) difference. One patient with MCI showed up during
follow-up to fulfill the criteria for vascular dementia and was not included in the comparison between pMCI and sMCI.

Microstates Results Between HC, MCI,
and AD
The average GEV was not significantly different between HC
(mean = 0.54, SD = 0.08), MCI (mean = 0.53, SD = 0.10), and AD
(mean = 0.56, SD = 0.03), (p-value = 0.3624, F-value = 1.0290).
See Figure 1 for global maps of the microstates that
were used for back-fitting. Between AD, MCI, and HC, we
found significantly different occurrence (p-value = 0.0277,
F-value = 3.7807, degrees of freedom = 68) and coverage
(p-value = 0.0101, F-value = 4.9237, degrees of freedom = 68)
for microstate A, see Table 4. For the post hoc t-test
for occurrence we found a significant difference between
AD, and HC (p-value = 0.0395, t-value = 2.1142, Cohen’s
d = 0.6471) and between MCI, and HC (p-value = 0.0411,
t-value = 2.0874, Cohen’s d = 0.5404). For coverage,
we found a significant difference between AD, and HC
(p-value = 0.0066, t-value = 2.8359, Cohen’s d = 0.8681) and
between MCI, and HC (p-value = 0.0077, t-value = 2.7575,
Cohen’s d = 0.7139).
For the syntax analysis, we found patients with MCI and AD
were significantly more likely to transition from microstates
C to A, and for AD from and D to A compared to HC
when only looking at the observed transition percentages

Frontiers in Neuroscience | www.frontiersin.org 5 June 2019 | Volume 13 | Article 563

Musaeus et al.
FIGURE 2 | Flow diagram showing the number of participants recruited and drop-out over 3 years. Figure reproduced from Musaeus et al. (2018b).
for each microstate separately, see Figure 3. However,
when we performed the permutation test between observed
and expected percentage of transitions, we did not find
any systematic deviation of transition from randomness
(p-value > 0.05). See Table 6 for observed and expected
percentage of transitions.
Microstates Between pMCI and sMCI
No significant differences (p-value < 0.05) were found
between pMCI and sMCI for duration, occurrence,
or coverage. The largest difference in duration was
found for microstate D between pMCI and sMCI
(see Table 5).
We also performed syntax analysis, but no significant
differences were found between pMCI and sMCI.
Frontiers in Neuroscience | www.frontiersin.org
EEG Microstates in MCI
Correlation
No significant correlations were found between coverage,
and occurrence for microstates A and amyloid, total tau,
phosphorylated tau, MMSE, or ACE.
DISCUSSION
In the current exploratory study, we found that patients with
MCI, and AD compared to HC had significantly higher
occurrence and coverage of microstate A. In addition,
both microstates C and D transitioned significantly more
to microstate A in patients with AD compared to HC,
and microstate C transitioned more to microstate A in
MCI compared to HC. However, we did not find evidence
6 June 2019 | Volume 13 | Article 563
Musaeus et al. EEG Microstates in MCI

p-value
that there was any systematic deviation of transition

0.578
0.307
0.01
0.26
probabilities from randomness for any of the groups.
Between pMCI, and sMCI, we did not find any significant
differences but the largest difference in duration was

0.24 (0.04)
0.21 (0.07)

0.29 (0.08)
0.26 (0.8)
found for microstate D. Lastly, no correlations were

AD
found between microstate A and either biomarkers or

TABLE 4 | Table showing the mean, standard deviation (SD), and p-value for comparisons between the three groups (HC, MCI, and AD) for microstates A–D for duration, occurrence, and coverage.
neuropsychological tests.

Coverage
Previous studies have investigated patients with AD (Ihl

0.23 (0.06)
0.24 (0.08)
0.25 (0.15)
0.28 (0.09)
et al., 1993; Dierks et al., 1997; Strik et al., 1997; Stevens and

MCI
Kircher, 1998; Nishida et al., 2013) but the majority found a
shorter duration of the microstates in patients suffering from
AD (Dierks et al., 1997; Strik et al., 1997; Stevens and Kircher,
1998) compared to healthy older controls. In the early studies

0.19 (0.06)
0.21 (0.09)
0.27 (0.11)
0.34 (0.13)
(Ihl et al., 1993; Dierks et al., 1997; Strik et al., 1997; Stevens

HC
and Kircher, 1998) adaptive segmentation was used, which may
have given rise to different results. However, a more recent study
using clustering analysis (Nishida et al., 2013) did not find any

p-value

0.028
0.422
0.256
0.304
significant differences between patients with AD and HC, which
could be due to low sample size or as previously suggested
temporal disorganization in patients with AD (Koenig et al.,
2005; Nishida et al., 2013). However, in the current study we

2.83 (0.45)
2.62 (0.68)
3.01 (0.56)
3.09 (0.41)
found longer duration in both patients with MCI, and AD with

AD
significant increased occurrence and coverage in microstate A

Occurrence
compared to HC, see Table 4. The main reason for our finding
compared to the recent study not finding any significant results

2.78 (0.55)
2.81 (0.67)
2.61 (0.59)
2.93 (0.66)
between AD, and HC (Nishida et al., 2013) may be differences

MCI
in methods. Here, we compiled GFP peaks for all participants
before segmentation or it could be due to differences in the
recruitment. However, significantly increased occurrence and

2.45 (0.65)
2.56 (0.87)
2.98 (0.76)
3.16 (0.63)
coverage for microstate A has not previously been reported

HC
and the underlying reason may be underlying AD pathology
in the temporal lobes, which has been shown in pathological
studies using Braak staging (Braak and Braak, 1991; Thal et al.,
2002) and studies using follow-up data on the deposition of
p-value

0.059
0.082
0.725
0.55
amyloid with PiB-PET (Okello et al., 2009; Villemagne et al.,
2011). The increased coverage and occurrence of microstate A
may therefore be due to underlying pathological changes in the
83.19 (11.05)

87.12 (16.95)
94.32 (22.69)
78.62 (9.15)

temporal lobes and thereby disruption of the underlying neuronal

networks. Interestingly, no statistically significant differences
AD

were found between MCI, and AD, which may indicate that the
majority of the included patients with MCI had an underlying
Duration

AD pathology. Supportive of this assumption, is the observation

82.23 (11.05)
85.45 (11.48)
90.76 (42.38)
92.26 (16.32)

that more than half of the MCI cohort progressed significantly

MCI

clinically over 2 years follow-up. However, we did not find

any significant changes in the microstate B between AD, MCI,
and HC, which may be due the topographical map did not
involve as large a region of the temporal region as microstate
75.88 (10.05)
78.10 (12.42)
89.08 (23.19)
106.02 (41.73)

A. However, in a previous paper using the same data for

spectral power analysis, we found that the changes are more
HC

pronounced on the left side (Musaeus et al., 2018b). This effect

may correspond to previous MR studies showing atrophy being
more pronounced on the left side (Killiany et al., 2000; Baron
Microstate C, (SD)
Microstate D, (SD)
Microstate B, (SD)
Microstate A, (SD)

et al., 2001) or a previous study showing more pronounced

hypometabolism in the left temporal region using SPECT (Hogh
et al., 2004). On the other hand, this may also simply be due
to low sample size and thereby individual differences affecting
the results. For the syntax analysis, we found that patients

Frontiers in Neuroscience | www.frontiersin.org 7 June 2019 | Volume 13 | Article 563

Musaeus et al. EEG Microstates in MCI

FIGURE 3 | Significant results for the syntax analysis between HC, MCI, and AD. The first column is for HC, second for MCI, and third for AD. The values represent
the percentage of times when microstates C, and D transitioned to the other microstates. The figure shows that both microstates C, and D were more likely to
transition to microstate A in patients with AD and in patients with MCI microstate C transitioned significantly more to microstates A.

TABLE 5 | Table showing the mean, standard deviation (SD), and p-value for comparisons between pMCI, and sMCI for microstates A-D for duration, occurrence,
and coverage.

Duration Occurrence Coverage

pMCI sMCI p-value pMCI sMCI p-value pMCI sMCI p-value

Microstate A, (SD) 75.32 (8.98) 76.22 (10.48) 0.594 2.41 (0.67) 2.47 (0.53) 0.587 0.18 (0.06) 0.19 (0.05) 0.547
Microstate B, (SD) 79.32 (16.66) 76.29 (8.49) 0.296 2.41 (0.96) 2.62 (0.84) 0.654 0.20 (0.12) 0.20 (0.08) 0.463
Microstate C, (SD) 81.59 (18.26) 95.06 (22.59) 0.673 2.61 (0.74) 3.31 (0.48) 0.708 0.22 (0.11) 0.31 (0.07) 0.655
Microstate D, (SD) 116.90 (33.91) 93.08 (26.15) 0.235 3.30 (0.58) 3.15 (0.75) 0.292 0.39 (0.14) 0.30 (0.10) 0.238

with both AD, and MCI were more likely to transition from
microstate C to A, and AD from D to A, see Figure 3.
However, we did not find any systematic deviation of transition
probabilities from randomness, which strongly indicates that the
transitions were in large part due to the increased occurrence
of microstate A.
The microstate classes have also been associated with BOLD
signal and resting state networks obtained with fMRI in multiple
studies (Britz et al., 2010; Van de Ville et al., 2010; Yuan
et al., 2012). One study has associated microstate A with
BOLD activations in the superior and middle temporal gyri
as well as the left middle frontal gyrus (Britz et al., 2010).
Other studies extracted 13 (Yuan et al., 2012) and 10
(Musso et al., 2010) microstates, respectively, and any direct
comparisons were therefore very difficult. By visual inspection,
it is possible that microstate A may correspond to microstate

Frontiers in Neuroscience | www.frontiersin.org 8 June 2019 | Volume 13 | Article 563

Musaeus et al. EEG Microstates in MCI

TABLE 6 | Observed and expected percentage of transitions.

HC MCI AD

Observed Expected Observed Expected Observed Expected

D to C 0.118 0.111 0.084 0.084 0.1 0.096

D to A 0.083 0.088 0.089 0.089 0.094 0.092
D to B 0.087 0.089 0.088 0.089 0.076 0.081
C to D 0.119 0.108 0.085 0.083 0.099 0.095
C to A 0.072 0.079 0.078 0.079 0.082 0.086
C to B 0.076 0.081 0.077 0.078 0.08 0.079
A to D 0.084 0.08 0.088 0.087 0.092 0.088
A to C 0.074 0.074 0.077 0.077 0.084 0.084
A to B 0.062 0.066 0.085 0.085 0.069 0.073
B to D 0.086 0.082 0.089 0.087 0.079 0.077
B to C 0.075 0.076 0.079 0.078 0.076 0.076
B to A 0.065 0.067 0.082 0.085 0.07 0.072

5 and 13 in a previous publication (Yuan et al., 2012)
and thereby be associated with the default mode network.
These findings suggest that microstate A is associated with
temporal connectivity and may even be related to the
default mode network.
In patients with pMCI and sMCI, we did not find
any significant changes but the largest difference in
duration was found for microstate D, see Table 5.
Microstate D has previously been associated with
BOLD changes in the frontal and parietal areas
measured with fMRI (Britz et al., 2010) and may
reflect underlying pathological changes in patients with
MCI who progress to AD. However, larger studies are
needed to test whether microstate D is in fact different
between pMCI and sMCI.
Previous studies have found an inverse correlation between
microstate lifespan and degree of cognitive impairment (Dierks
et al., 1997; Strik et al., 1997). In the current study, we
did not find any correlation between occurrence or coverage
and either biomarkers or neuropsychological tests. This may
be due to the low sample size or the values extracted
based on the global maps. Larger studies are needed to
investigate whether microstate changes are associated with
neuropsychological findings.
In the current study, we choose to extract four microstates
since this is the most commonly reported and these have
been shown to be reliable (Khanna et al., 2014). However,
the GEV was not significantly different between the three
groups but was low (average GEV = 54%) compared to
other studies with most commonly reporting a GEV >70%
(Michel and Koenig, 2018). The low GEV may be due
to either broad filter settings (2–20 Hz) or simply due to
patient data being noisier. In the current analysis, we included
only the first 1000 GFP peaks to the segmentation and
thereby avoided problems in terms of more contributions from
larger EEG files.
The study indicates that microstate A could be an early disease
marker in patients with MCI, but it has some limitations. Firstly,
we acknowledge the relatively small sample size and we did
not correct for multiple comparisons due to the exploratory
nature of the study. However, these changes suggest that
larger studies will be able to use microstates as a classifier
of disease even at an early stage. In addition, the follow-
up time was short and according to previous studies, annual
clinical progression rate is 15% (Petersen et al., 1999; Saxton
et al., 2009), which means that only 30% of the patients
with MCI should have progressed to AD. However, we found
that 48% progressed, which may in part be due to the
patients with MCI being at a more advanced stage of the
disease at inclusion. Furthermore, we included patients receiving
medication in the analysis, which may have affected the EEG.
Nevertheless, our findings in this small pilot study with affected
microstate A in patients with MCI and possible affection of
microstate D in the transition from MCI to AD may be able to
guide larger studies.
CONCLUSION
In the current exploratory study, we found that patients
with MCI, and AD compared to HC had significantly
higher occurrence and coverage of microstate A. The
changes may correspond to the previous literature of
pathological changes in the temporal regions in patients
with AD and microstate A may correspond to temporal
regions measured with BOLD fMRI. Furthermore, between
pMCI, and sMCI, no significant differences were found
but a tendency of a prolonged duration of microstate D in
patients with pMCI was seen. Larger studies are needed to
confirm these findings.
DATA AVAILABILITY
The datasets supporting the conclusions of this manuscript
will be made available by the authors to any qualified
researcher. However, due to regulations, we are not able to
share the EEG files.

Frontiers in Neuroscience | www.frontiersin.org 9 June 2019 | Volume 13 | Article 563

Musaeus et al.
ETHICS STATEMENT
This study was carried out in accordance with the
recommendations of the Regional Committee on Health
Research Ethics with written informed consent from all subjects.
All subjects gave written informed consent in accordance with
the Declaration of Helsinki. The protocol was approved by the
Regional Committee on Health Research Ethics.
AUTHOR CONTRIBUTIONS
PH, MN, and CM conceived the project idea of
using quantitative EEG. PH and MN conducted the
REFERENCES
Baron, J. C., Chetelat, G., Desgranges, B., Perchey, G., Landeau, B., de la Sayette,
V., et al. (2001). In vivo mapping of gray matter loss with voxel-based
morphometry in mild Alzheimer’s disease. NeuroImage 14, 298–309. doi: 10.
1006/nimg.2001.0848
Braak, H., and Braak, E. (1991). Neuropathological stageing of Alzheimer-related
changes. Acta Neuropathol. 82, 239–259. doi: 10.1007/bf00308809
Britz, J., Van De Ville, D., and Michel, C. M. (2010). BOLD correlates of
EEG topography reveal rapid resting-state network dynamics. NeuroImage 52,
1162–1170. doi: 10.1016/j.neuroimage.2010.02.052
Chui, H. C. (1987). The significance of clinically defined subgroups of Alzheimer’s
disease. J. Neural Transm. Suppl. 24, 57–68.
Cummings, J. L. (2004). Alzheimer’s disease. N. Engl. J. Med. 351, 56–67.
Delorme, A., and Makeig, S. (2004). EEGLAB: an open source toolbox for
analysis of single-trial EEG dynamics including independent component
analysis. J. Neurosci. Methods 134, 9–21. doi: 10.1016/j.jneumeth.2003.
10.009
Dickerson, B. C., and Sperling, R. A. (2009). Large-scale functional brain network
abnormalities in Alzheimer’s disease: insights from functional neuroimaging.
Behav. Neurol. 21, 63–75. doi: 10.3233/BEN-2009-0227
Dierks, T., Jelic, V., Julin, P., Maurer, K., Wahlund, L. O., Almkvist, O., et al.
(1997). EEG-microstates in mild memory impairment and Alzheimer’s disease:
possible association with disturbed information processing. J. Neural Transm.
104, 483–495. doi: 10.1007/bf01277666
Engedal, K., Snaedal, J., Hoegh, P., Jelic, V., Bo Andersen, B., Naik, M., et al.
(2015). Quantitative EEG applying the statistical recognition pattern method:
a useful tool in dementia diagnostic workup. Dement. Geriatr. Cogn. Dis. 40,
1–12. doi: 10.1159/000381016
Hogh, P., Madsen Sjo, N., Gade, A., and Waldemar, G. (2004). Temporal lobe
hypoperfusion in isolated amnesia with slow onset: a single photon emission
computer tomography study. Dement. Geriatr. Cogn. Dis. 18, 15–23. doi: 10.
1159/000077730
Ihl, R., Dierks, T., Froelich, L., Martin, E. M., and Maurer, K. (1993). Segmentation
of the spontaneous EEG in dementia of the Alzheimer type. Neuropsychobiology
27, 231–236. doi: 10.1159/000118986
Khanna, A., Pascual-Leone, A., and Farzan, F. (2014). Reliability of resting-state
microstate features in electroencephalography. PLoS One 9:e114163. doi: 10.
1371/journal.pone.0114163
Khanna, A., Pascual-Leone, A., Michel, C. M., and Farzan, F. (2015). Microstates in
resting-state EEG: current status and future directions. Neurosci. Biobehav. Rev.
49, 105–113. doi: 10.1016/j.neubiorev.2014.12.010
Killiany, R. J., Gomez-Isla, T., Moss, M., Kikinis, R., Sandor, T., Jolesz, F.,
et al. (2000). Use of structural magnetic resonance imaging to predict who
will get Alzheimer’s disease. Ann. Neurol. 47, 430–439. doi: 10.1002/1531-
8249(200004)47:4<430::aid-ana5>3.3.co;2-9
Koenig, T., Prichep, L., Lehmann, D., Sosa, P. V., Braeker, E., Kleinlogel, H., et al.
(2002). Millisecond by millisecond, year by year: normative EEG microstates
and developmental stages. NeuroImage 16, 41–48. doi: 10.1006/nimg.2002.1070
Frontiers in Neuroscience | www.frontiersin.org
EEG Microstates in MCI
experiments. CM conducted the data analyses and drafted
the manuscript. PH, MN, and CM contributed to
revising the manuscript.
FUNDING
This work was funded by the Velux Foundation.
ACKNOWLEDGMENTS
We would like to thank the study nurses for all their help
conducting this study.
Koenig, T., Studer, D., Hubl, D., Melie, L., and Strik, W. K. (2005). Brain
connectivity at different time-scales measured with EEG. Philos. Trans. R. Soc.
Lond. B Biol. Sci. 360, 1015–1023.
Lee, T. W., Girolami, M., and Sejnowski, T. J. (1999). Independent component
analysis using an extended infomax algorithm for mixed subgaussian
and supergaussian sources. Neural Comput. 11, 417–441. doi: 10.1162/
089976699300016719
Lehmann, D. (1971). Multichannel topography of human al pha EEG fields.
Electroencephalogr. Clin. Neurophysiol. 33, 439–449. doi: 10.1016/0013-
4694(71)90165-9
Lehmann, D., Faber, P. L., Galderisi, S., Herrmann, W. M., Kinoshita, T., Koukkou,
M., et al. (2005). EEG microstate duration and syntax in acute, medication-
naive, first-episode schizophrenia: a multi-center study. Psychiatry Res. 138,
141–156. doi: 10.1016/j.pscychresns.2004.05.007
Lehmann, D., Ozaki, H., and Pal, I. (1987). EEG alpha map series: brain
micro-states by space-oriented adaptive segmentation. Electroencephalogr. Clin.
Neurophysiol. 67, 271–288. doi: 10.1016/0013-4694(87)90025-3
McKhann, G. M., Knopman, D. S., Chertkow, H., Hyman, B. T., Jack, C. R. Jr.,
Kawas, C. H., et al. (2011). The diagnosis of dementia due to Alzheimer’s disease:
recommendations from the national institute on aging-alzheimer’s association
workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers
Dement. 7, 263–269. doi: 10.1016/j.jalz.2011.03.005
Michel, C. M., and Koenig, T. (2018). EEG microstates as a tool for studying the
temporal dynamics of whole-brain neuronal networks: a review. NeuroImage
180(Pt B), 577–593. doi: 10.1016/j.neuroimage.2017.11.062
Musaeus, C. S., Engedal, K., Hogh, P., Jelic, V., Morup, M., Naik, M., et al. (2018a).
EEG theta power is an early marker of cognitive decline in dementia due to
alzheimer’s disease. J. Alzheimers Dis. 64, 1359–1371. doi: 10.3233/JAD-180300
Musaeus, C. S., Nielsen, M. S., Osterbye, N. N., and Hogh, P. (2018b). Decreased
parietal beta power as a sign of disease progression in patients with mild
cognitive impairment. J. Alzheimers Dis. 65, 475–487. doi: 10.3233/JAD-180384
Musaeus, C. S., Nielsen, M. S., and Hogh, P. (2019). Altered low-frequency EEG
connectivity in mild cognitive impairment as a sign of clinical progression.
J. Alzheimers Dis. 68, 947–960. doi: 10.3233/JAD-181081
Musso, F., Brinkmeyer, J., Mobascher, A., Warbrick, T., and Winterer, G. (2010).
Spontaneous brain activity and EEG microstates. A novel EEG/fMRI analysis
approach to explore resting-state networks. NeuroImage 52, 1149–1161.
doi: 10.1016/j.neuroimage.2010.01.093
Nielsen, M. S., Simonsen, A. H., Siersma, V., Hasselbalch, S. G., and Hoegh, P.
(2018). The diagnostic and prognostic value of a dual-tasking paradigm in a
memory clinic. J. Alzheimers Dis. 61, 1189–1199. doi: 10.3233/JAD-161310
Nishida, K., Morishima, Y., Yoshimura, M., Isotani, T., Irisawa, S., Jann, K.,
et al. (2013). EEG microstates associated with salience and frontoparietal
networks in frontotemporal dementia, schizophrenia and Alzheimer’s disease.
Clin. Neurophysiol. 124, 1106–1114. doi: 10.1016/j.clinph.2013.01.005
Okello, A., Koivunen, J., Edison, P., Archer, H. A., Turkheimer, F. E., Nagren, K.,
et al. (2009). Conversion of amyloid positive and negative MCI to AD over
3 years: an 11C-PIB PET study. Neurology 73, 754–760. doi: 10.1212/WNL.
0b013e3181b23564
10 June 2019 | Volume 13 | Article 563
Musaeus et al.
Oostenveld, R., Fries, P., Maris, E., and Schoffelen, J. M. (2011). FieldTrip:
open source software for advanced analysis of MEG, EEG, and invasive
electrophysiological data. Comput. Intell. Neurosci. 2011:156869. doi: 10.1155/
2011/156869
Pascual-Marqui, R. D., Michel, C. M., and Lehmann, D. (1995). Segmentation of
brain electrical activity into microstates: model estimation and validation. IEEE
Trans. Biomed. Eng. 42, 658–665. doi: 10.1109/10.391164
Petersen, R. C. (2004). Mild cognitive impairment as a diagnostic entity. J. Int. Med.
256, 183–194. doi: 10.1111/j.1365-2796.2004.01388.x
Petersen, R. C., Smith, G. E., Waring, S. C., Ivnik, R. J., Tangalos, E. G., and
Kokmen, E. (1999). Mild cognitive impairment: clinical characterization and
outcome. Arch. Neurol. 56, 303–308.
Poulsen, A. T., Pedroni, A., Langer, N., and Hansen, L. K. (2018). Microstate
EEGlab toolbox: an introductory guide. bioRxiv
Saxton, J., Snitz, B. E., Lopez, O. L., Ives, D. G., Dunn, L. O., Fitzpatrick,
A., et al. (2009). Functional and cognitive criteria produce different rates of
mild cognitive impairment and conversion to dementia. J. Neurol. Neurosurg.
Psychiatry 80, 737–743. doi: 10.1136/jnnp.2008.160705
Selkoe, D. J. (2002). Alzheimer’s disease is a synaptic failure. Science 298, 789–791.
Stevens, A., and Kircher, T. (1998). Cognitive decline unlike normal aging is
associated with alterations of EEG temporo-spatial characteristics. Eur. Arch.
psychiatry Clin. Neurosci. 248, 259–266. doi: 10.1007/s004060050047
Strik, W. K., Chiaramonti, R., Muscas, G. C., Paganini, M., Mueller, T. J., Fallgatter,
A. J., et al. (1997). Decreased EEG microstate duration and anteriorisation of
the brain electrical fields in mild and moderate dementia of the Alzheimer type.
Psychiatry Res. 75, 183–191. doi: 10.1016/s0925-4927(97)00054-1
Thal, D. R., Rub, U., Orantes, M., and Braak, H. (2002). Phases of a beta-deposition
in the human brain and its relevance for the development of AD. Neurology 58,
1791–1800. doi: 10.1212/wnl.58.12.1791
Frontiers in Neuroscience | www.frontiersin.org
EEG Microstates in MCI
Van de Ville, D., Britz, J., and Michel, C. M. (2010). EEG microstate sequences in
healthy humans at rest reveal scale-free dynamics. Proc. Natl. Acad. Sci. U.S.A.
107, 18179–18184. doi: 10.1073/pnas.1007841107
Villemagne, V. L., Pike, K. E., Chetelat, G., Ellis, K. A., Mulligan, R. S., Bourgeat,
P., et al. (2011). Longitudinal assessment of abeta and cognition in aging and
Alzheimer disease. Ann. Neurol. 69, 181–192.
Wackermann, J., Lehmann, D., Michel, C. M., and Strik, W. K. (1993).
Adaptive segmentation of spontaneous EEG map series into spatially defined
microstates. Int. J. Psychophysiol. 14, 269–283. doi: 10.1016/0167-8760(93)
90041-m
Winblad, B., Palmer, K., Kivipelto, M., Jelic, V., Fratiglioni, L., Wahlund, L. O., et al.
(2004). Mild cognitive impairment–beyond controversies, towards a consensus:
report of the international working group on mild cognitive impairment.
J. Intern. Med. 256, 240–246.
Yuan, H., Zotev, V., Phillips, R., Drevets, W. C., and Bodurka, J. (2012).
Spatiotemporal dynamics of the brain at rest–exploring EEG microstates as
electrophysiological signatures of BOLD resting state networks. NeuroImage 60,
2062–2072. doi: 10.1016/j.neuroimage.2012.02.031
Conflict of Interest Statement: The authors declare that the research was
conducted in the absence of any commercial or financial relationships that could
be construed as a potential conflict of interest.
Copyright © 2019 Musaeus, Nielsen and Høgh. This is an open-access article
distributed under the terms of the Creative Commons Attribution License (CC BY).
The use, distribution or reproduction in other forums is permitted, provided the
original author(s) and the copyright owner(s) are credited and that the original
publication in this journal is cited, in accordance with accepted academic practice. No
use, distribution or reproduction is permitted which does not comply with these terms.
11 June 2019 | Volume 13 | Article 563